Conventionally in an HEV, the state of charge (SOC) of a secondary battery is estimated by computation from the detected voltage, current, temperature and other aspects of the secondary battery, and then SOC control is performed in order to optimize the fuel efficiency of the vehicle. In order to perform SOC control accurately, the SOC of the secondary battery must be estimated accurately as it is discharged and charged.
A conventional method for performing this SOC estimation is carried out as follows. First, the battery voltage V and the discharge or charge current I are measured over a predetermined period of time, and the integral ∫I of that current is calculated. In addition, a temperature T, a battery voltage V, and a function of the current integral ∫I are used to update the previous estimate of the polarization voltage of the cell Vc(t−1) to Vc(t), and find a correction voltage V′ (=V−Vc(t)), and then a plurality of pairs of data, each pair consisting of the correction voltage V′ and the current I, are obtained and stored. From these pairs of data, regression analysis is used to find a first-order straight-line approximation (voltage V′-current I straight-line approximation), and the V intercept of the V′-I straight-line approximation is estimated to be the electromotive force E. The SOC is then estimated from the previous estimate of SOC, electromotive force E, the temperature T and the function of the current integral ∫I (e.g., JP 2001-223033A).
However, the conventional method for estimating SOC described above has the following problems.
First of all, the discharging or charging current flowing through the secondary battery is measured by a current sensor in order to estimate the SOC. When used in a HEV or the like, this current sensor must measure a large current and using a high-precision sensor would mean increased cost, so that the practical situation is such that it is necessary to use a low-cost and relatively imprecise sensor. For this reason, the current value detected by the current sensor contains a measurement error and this current error results in an error in the estimation of the SOC. Particularly in cases in which the rate of discharging or charging is smaller than the current error (cases such as when the current error is ±2 A compared to a rate of discharging or charging of 1 A), the behavior of the estimated SOC becomes extremely strange.
In addition, in the conventional method for estimating the SOC described above wherein the previous estimate of the polarization voltage of the cell Vc(t−1) is updated to Vc(t) as a function of the integral of the current measured by this current sensor so that the effect of the polarization voltage is taken into consideration, the previously calculated polarization voltage contains a current error, so that this current error becomes an error in the estimation of the polarization voltage and these errors accumulate, resulting in a problem wherein the error between the estimated value and the true value of the SOC becomes greater with the passage of time.